Start with the mission
Choose a concrete challenge where better information, faster workflows, or more consistent support could create meaningful value. Define who benefits, what success looks like, and which decisions must remain with people. Start small enough to learn quickly, while keeping the organization’s values and community commitments visible throughout the work.Discover a useful starting point
Explore Public Libraries and your organization’s approved libraries before creating something new. Review a resource’s purpose, inputs, connected tools, data handling, and expected outputs. Reuse or fork a starting point only when it fits your mission, your communities, and your team’s capacity to support it.Build the right kind of tool
- Prompts help staff apply a clear, repeatable instruction to a common task. Create a Prompt.
- Blueprints connect AI, knowledge, data, tools, and human review into a visible workflow. Create a Blueprint.
- Chatbots can provide a guided experience when a conversation is the right way to support staff, volunteers, partners, or community members. Create a Chatbot.
- Agents can bring together a defined role, approved knowledge, tools, and boundaries for a reusable assistant.
Test with care and learn from people
Use de-identified, synthetic, or otherwise approved test data whenever possible. Test ordinary requests, incomplete information, and situations where an answer could be confusing, harmful, or unfair. Include the staff and community perspectives closest to the work, and make it easy for users to spot limitations, question an output, or ask a person for help.Build trust before you scale
Before sharing a tool more broadly, confirm its purpose, accountable owner, access permissions, data practices, human-review points, and communication plan. Consider accessibility, language, privacy, security, and digital equity alongside model quality. Keep a tool private while it is being refined, and share only the version your organization is ready to support.A simple AI-for-Good cycle
- Identify a mission need and the people it should benefit.
- Discover a reusable pattern or build a small prototype.
- Test with safe data and representative users.
- Review impact, risks, and governance with the right people.
- Improve, share, or scale what proves useful—and document what the team learned.